Automated Retrieval of ACLED Conflict Event Data

CRAN status Lifecycle: stable Travis build status License: CC BY-NC 4.0 CRANlogs

This small package provides functionality to access and manage the application programming interface (API) of the Armed Conflict Location & Event Data Project (ACLED), while requiring a minimal number of dependencies. The function acled.api() makes it easy to retrieve a user-defined sample (or all of the available data) of ACLED, enabling a seamless integration of regular data updates into the research work flow.

When using this package, you acknowledge that you have read ACLED’s terms and conditions of use, and that you agree with their attribution requirements.

Installation

You can install the latest release version of acled.api from CRAN with:

install.packages("acled.api") # downloads and installs the package from CRAN

You can install the development version from GitLab with:

remotes::install_gitlab("chris-dworschak/acled.api") # downloads and installs the package from GitLab

Example

Using acled.api is straight forward. To download data on, for example, all ACLED conflict events in Europe and Central America that happened between June 2019 and July 2020, you can supply:

library(acled.api) # loads the package
#> 
#> By using this package, you acknowledge that you have read ACLED's terms and
#> conditions. The data must be cited as per ACLED attribution requirements. To
#> download ACLED data, you require an ACLED access key. You can request your key
#> by freely registering with ACLED on https://developer.acleddata.com/.
#> The package may be cited as:
#> Dworschak, Christoph. 2020. "Acled.api: Automated Retrieval of ACLED Conflict
#> Event Data." R package. CRAN version 1.1.4.
#> For the development version of this package, visit <https://gitlab.com/chris-dworschak/acled.api/>
my.data.frame <- acled.api( # stores an ACLED sample in object my.data.frame
  email.address = Sys.getenv("EMAIL_ADDRESS"),
  access.key = Sys.getenv("ACCESS_KEY"),
  region = c("South Asia", "Central America"), 
  start.date = "2019-09-01", 
  end.date = "2020-01-31")
#> Your ACLED data request was successful. 
#> Events were retrieved for the period starting 2019-09-01 until 2020-01-31.
my.data.frame[1:5,] # returns the first three observations of the ACLED sample
#>       region country year event_date                    source      admin1
#> 1 South Asia   India 2020 2020-01-31 India Blooms News Service       Delhi
#> 2 South Asia   India 2020 2020-01-31         Telegraph (India) West Bengal
#> 3 South Asia   India 2020 2020-01-31                 The Hindu Maharashtra
#> 4 South Asia   India 2020 2020-01-31         Telegraph (India) West Bengal
#> 5 South Asia   India 2020 2020-01-31                 The Hindu  Tamil Nadu
#>             admin2          admin3            location event_type
#> 1       East Delhi     Preet Vihar Delhi - Laxmi Nagar   Protests
#> 2          Kolkata         Kolkata             Kolkata   Protests
#> 3  Mumbai Suburban Mumbai Suburban              Mumbai   Protests
#> 4 Dakshin Dinajpur     Gangarampur         Gangarampur      Riots
#> 5            Salem           Salem               Salem      Riots
#>              sub_event_type interaction fatalities  timestamp
#> 1 Protest with intervention          16          0 1618511523
#> 2          Peaceful protest          60          0 1580841823
#> 3          Peaceful protest          60          0 1618511527
#> 4              Mob violence          57          0 1581453825
#> 5     Violent demonstration          50          0 1580841823

A note on replicability

Some tasks, like real-time analyses and continuously updated forecasting models (e.g., as used by practitioners), may not require replicability of results. However, most research-related tasks assume the possibility of replication at a later stage (e.g., when results are intended for publication, or a data project taking multiple days where a change to the underlying sample is not desirable). After the release of versions 1 through 8, ACLED changed their update system to allow for real-time amendments and post-release corrections, thereby forgoing traditional data versioning. This change requires researchers to take additional steps in order to ensure the replicability of their results when using ACLED data.

Importantly, downloaded data intended for replicable use must be permanently stored by the analyst. Data downloaded through acled.api() are only stored temporarily in the working space, and may be lost after closing R. Therefore, if replicability is important to the analyst’s task, a call through acled.api() should occur only once at the beginning of the data project, immediately followed by, e.g., saveRDS(downloaded.data, file = "my_acled_data.rds"). This locally stored data file can then be used again at a later point by calling readRDS(file = "my_acled_data.rds"), and ensures that the analysis sample stays constant over time.

ACLED provides a time stamp for each individual observation, enabling researchers to do “micro versioning” of data points if necessary, and to verify congruence across samples. For this it is important that researchers do not drop the variable timestamp during the data management process. Starting version 1.0.9 the function acled.api() includes the timestamp variable in its default API call.